Researchers say chatbots copy human speech habits but lack human creativity

In short: Chatbots are getting better at sounding human in everyday conversation, but researchers say they still do not have human, experience-based creativity or judgment.

What's going on

Recent research in communication and ethics describes modern chatbots as “artificial rhetorical agents.” That means they can argue, take on a role like “helpful expert,” and switch tone on request. But their style comes from training on large amounts of human writing, so they mostly remix patterns they have seen before, like a very fast autocomplete.

Studies also show these systems are good at copying surface-level speech habits. They can match emotional framing, use polite filler phrases, and hedge with language like “it depends” or “I might be wrong.” In group-like settings, they can even drift toward similar views, which looks like people copying each other in a crowd.

At the same time, consumer-protection and AI-ethics writers warn that this human-like voice can create “counterfeit people.” The chatbot can sound caring or humble, but it does not feel anything and it does not have anything at stake. Experts say this can lead users to trust a bot too much, especially in areas like advertising, persuasion, and politics.

What to watch

More researchers and product teams are now treating “how a bot talks” as a design choice, not just a side effect. That includes adding more warnings, refusals, and cautious language. For regular users, the key skill may be learning to ask better questions and to double-check answers, treating chatbot replies like drafts from a stranger, not advice from a responsible person.

Source: NYTimes